geosciences Communication
How Characterization of Particle Size Distribution Pre- and Post-Reaction Provides Mechanistic Insights into Mineral Carbonation † Aashvi Dudhaiya and Rafael M. Santos *
ID
School of Engineering, University of Guelph, Guelph, ON N1G 2W1, Canada;
[email protected] * Correspondence:
[email protected]; Tel.: +1-519-824-4120 (ext. 52902) † Some contents of this paper also appear in the Proceedings of the 8th World Congress on Particle Technology Meeting (Orlando, USA, 2018), as extended abstract P508057, titled “Particle size characterization in mineral carbonation for understanding reaction fundamentals”. Received: 22 June 2018; Accepted: 9 July 2018; Published: 11 July 2018
Abstract: Mineral carbonation is the conversion of carbon dioxide, in gas form or dissolved in water, to solid carbonates. Materials characterization plays an important role in assessing the potential to use these carbonates in commercial applications, and also aids in understanding fundamental phenomena about the reactions. This paper highlights findings of mechanistic nature made on topics related to mineral carbonation, and that were made possible by assessing particle size, particle size distribution, and other morphological characteristics. It is also shown how particle size data can be used to estimate the weathering rate of carbonated minerals. An extension of the carbonation weathering rate approach is presented, whereby using particle size distribution data it becomes possible to predict the particle size below which full carbonation is obtained, and above which partial carbonation occurs. The paper also overviews the most common techniques to determine the particle size distribution, as well as complementary and alternate techniques. In mineral carbonation research, most techniques have been used as ex situ methods, yet tools that can analyze powders during reaction (in situ and real-time) can provide even more insight into mineral carbonation mechanisms, so researchers are encouraged to adopt such advanced techniques. Keywords: mineral carbonation; particle size distribution; laser diffraction; carbonation conversion; iron- and steel-making slags; carbonation weathering rate
1. Introduction Mineral carbonation is the conversion of carbon dioxide, in gas form or dissolved in water, to solid carbonates. These may include calcium carbonates, magnesium carbonates, and a variety of other alkaline earth metal carbonates. Alkaline earth metals can be derived from natural minerals, waste residues, or even brines [1]. When performed in a reactor, mineral carbonation leads to the formation of a powder that can be composed of a carbonate alone (e.g., high-purity precipitated calcium carbonate) [2], or a mixture of one or more carbonates with other solid phases, such as silica and silicates [3]. Materials characterization thus plays an important role in assessing the potential to use these carbonates in commercial applications (e.g., filler in paper, aggregate in concrete) [4]. Materials characterization also aids in understanding fundamental phenomena about the reactions, such as the behavior of the reagents, the formation of products and by-products, and the reaction rate and conversion limitations [5]. A materials characterization parameter regularly measured and reported is the particle size. This can include the particle size of the starting materials, or that of the converted minerals, at the end
Geosciences 2018, 8, 260; doi:10.3390/geosciences8070260
www.mdpi.com/journal/geosciences
Geosciences 2018, 8, 260
2 of 20
of the reaction or as a function of the reaction extent. Particle size is expressed as an average value, as an interval, or as a distribution. Particle size data is used to predict the susceptibility of a mineral powder to carbonation, or to explain the mineral carbonation conversion achieved, or with eye towards materials applications. While experimental research on mineral carbonation can be conducted with little to no use of particle size characterization (though at a loss of mechanistic understanding of the reaction), knowledge of particle size is essential to the modeling of mineral carbonation reactions, an active area of research in recent years [6–9]. Gopinath and Mehra [6] modeled steel slag carbonation by assuming particles had a slab geometry, and used particle size information to describe not only the original and post-carbonation slab length, but also the location within the slab of the reaction front. They, notably, pointed out that when particle size distribution data is available, the conversion can be computed as a weighted average of the conversion in each bin, whereas when the distribution data is not available, the mean particle diameter is the next best data required. Pan et al. [7] reviewed several models to describe solid-liquid mineral carbonation reactions (shrinking core model, surface coverage model, integrated reaction-diffusion model, two-layer diffusion model, integrated transport-reaction model), all of which require information about particle size. The traditional shrinking core model can account for changes in particle size due to reaction products, using “Z factor” and “λ factor”, but the modified shrinking core model reduces back to its original form when the particle size remains unchanged in the course of reaction [7,8]. The surface coverage model uses specific surface area of the reacting powder [9], which is related to the particle size for materials with low porosity and spheroidal shape (e.g., most slags), but can deviate significantly otherwise (e.g., fly ash), requiring dedicated determination. This paper highlights (Section 3.1) the findings of the mechanistic nature that the corresponding author has made on topics related to mineral carbonation, and that were made possible by assessing particle size, particle size distribution, and other morphological characteristics. Similar applications of particle size characterization are also reviewed (Section 3.2) from recent works related to the carbonation of iron- and stainless-steel slags. Subsequently (Section 4.1), it is shown how particle size data can be used to estimate the weathering rate (µm·min–1 ) of carbonated minerals, which is a measure that normalizes carbonation conversion data with respect to volume-based average particle size, thus, making it possible to compare the carbonation of different materials using a particular process, or the carbonation of similar materials using different processes on an equal basis [10]. The carbonation weathering rate is also ideal for assessing the intensification efficacy of advanced carbonation methodologies [10,11]. An extension of the CWR approach is presented in Section 4.2, whereby using particle size distribution data it becomes possible to predict the particle size below which full carbonation is obtained, and above which partial carbonation occurs. The paper begins (Section 2) with an overview of the most common techniques to determine particle size distribution, as well as complementary and alternate techniques that provide additional particle size information and/or are able to record the particle size distribution in situ and under real process conditions. 2. Common, Complementary, and Alternate Particle Size Distribution Methodologies 2.1. Sieving Analysis Sieving is classified as a mechanical method for particle size distribution (PSD) determination, and is the simplest and most often used method in minerals research. Sieving relies on the use of screens with specified mesh sizes to separate fractions of a (typically) dry powder sample. Particles pass through the sieve with the aid of vibrations, which can be manually generated, or using a vibratory shaker. The goal of these vibrations is to allow the smaller particles to travel downwards, by gravity, through the inter-particles spacing (porosity) around the larger particles. To achieve a high degree of separation (i.e., fractionation), the most important parameters during sieving are amplitude (e.g., 1.5 mm) and time (e.g., 5 min) of shaking, as well as having the correct amount of sample loaded onto the coarsest sieve [12].
Geosciences 2018, 8, 260
3 of 20
Some important limitations of sieving reduce the value of this method for mineral carbonation research. One is that the smallest standard sieve size (400 mesh, or 0.036–0.038 mm) is significantly coarser than mineral particles often use for carbonation, especially when minerals are milled. The result is that researchers only know the passing fraction (e.g., 90 wt % passing 400 mesh), but not the exactle particle size distribution. Likewise, with coarser particles, it may be determined that a certain mass fraction of sample is between two mesh sizes in diameter (e.g., 25 wt % between 0.15 and 0.25 mm), but it is not possible to determine an exact value of average particle size even in this range, as the median value will only provide a rough estimate. Another important issue with sieving, related to mineral carbonation research, is that fine particles can be trapped within coarser fractions, as a result of poor separation due to particle shape or electrostatic forces. This can skew the particle size distribution mass fractions determined but, most importantly, can affect experimental results. If a material contains fine particles, those particles will carbonate faster than the coarser particles, affecting the determination of reaction kinetics, especially with short carbonation durations. Wet sieving (with water or non-aqueous solvent) is a technique that can reduce fine particle entrapment, but can also lead to alteration of the particles (e.g., partial dissolution, hydration, caking) depending on the mineral. Particle shape can also have an effect on sieving analysis results. The passage or the retention of a non-spherical particles, through or by a sieve of a given mesh size, is determined by the probability of the particle to assume an orientation relative to the mesh that allows it to pass through [13]; such an orientation exists when the particle’s smallest cross-section is smaller than the sieve’s aperture. The net outcome of the non-sphericity is that a coarser population is retained by a sieve than the actual population of particles with apparent diameters that match the sieve size [13]. Sieving, nonetheless, remains a valid laboratory technique as it can not only provide semi-quantitative particle size distribution information, but is also a solids separation technique. Mineral samples that are crushed, milled, sedimented, centrifuged, or precipitated can be passed through sieves to obtain a specific particle size fraction to be used for mineral carbonation reactions. These fractions can then be tested using advanced characterization techniques to accurately determine particle size distribution, average particle size, and particle shape. 2.2. Laser Diffraction Analysis Laser diffraction analysis (LDA) is a technique that is useful to follow particle size growth or reduction. Thus, it can help elucidate effects of carbonation, such as the formation of precipitated layers or alteration of the microstructure, and of sonication (a technique to intensify the carbonation reaction), such as particle attrition and fragmentation. Of the three average particle diameters commonly obtained by laser diffraction (D50, D(4,3), and D(3,2)), the Sauter mean diameter (D(3,2)), being surface area sensitive, is the most important when it comes to the susceptibility towards mineral carbonation of powdery materials, since mineral carbonation reactivity is proportional to the exposed surface area. This mean diameter is also found to be the most sensitive to sonication effects, as it best detects the formation of micron- to sub-micron sized fragments. The volume-moment mean diameter (D(4,3)), on the other hand, is useful in indicating the degree of erosion of larger particles. These insights are discussed in more detail in Section 3.1. The principle of laser diffraction analysis (LDA) is that particles of a given size diffract light through a given angle, which is inversely proportional to particle size, and the intensity of the diffracted beam at any angle is a measure of the number of particles with a specific cross-sectional area in the optical path [14]. For calculating particle sizes from the diffraction data, two techniques are commonly used: Fraunhofer diffraction and Mie theory. Both assume that the particles have a spherical shape (i.e., the particle dimension is the optical spherical diameter), but the Mie theory is deemed more accurate for clay-sized particles, hence, it is typically applied in mineral carbonation research [14]. Since laser diffraction inherently determines particle count, it is readily possible to present PSD data on a volume, surface area, or number basis, with the first being the most commonly reported and interpreted.
Geosciences 2018, 8, 260
4 of 20
Determination of the particle size distribution (PSD) by LDA offers a number of advantages over sieving (and sedimentation, another long-established mechanical technique used in soil analysis). Most useful is its ability to generate a complete particle size distribution (i.e., a particle diameter histogram with small bin size), which, according to Fisher et al. [15], is important for: (1) more completely quantifying differences between samples; (2) developing mathematical functions to describe and utilize the PSD; and (3) transferring data between different particle size classification systems. LDA is also often cited as being a superior method for accurately determining particle diameter (i.e., closest to true value), and for more precisely differentiating between similar samples [16,17]. PSD data is also directly used to calculate average particle sizes, such as the volume median diameter (D50), the volume moment mean diameter (D(4,3)), and the surface area moment (Sauter) mean diameter (D(3,2)), or cumulative percentiles such as D10 or D90. The operation of laser diffraction analyzer is best described by looking at the experimental procedure used in a specific study. Fisher et al. [15], in their study to compare LDA with the sedimentation-based sieve plummet balance technique, utilized a Malvern Mastersizer 2000 (Malvern Instruments, Malvern, UK) to carry out LDA, which uses a 52-detector array. Measurements were performed at a constant temperature of 20 ◦ C (± 2 ◦ C), and the samples were suspended and recirculated in water, with the pumping speed, stirring speed, and ultrasonic level set at 2000 rpm, 800 rpm, and 100%, respectively. Particle size was calculated on a volume basis using the Mie theory and Malvern proprietary software (version 5.6, Malvern Panalytical, Malvern, UK). The background measurements were taken for approximately 30 s, followed by measuring the suspended sample aliquot eight times (for replication) using a 30 s measurement duration each time, which is equivalent to 30,000 individual light scattering measurements. There are several factors that can affect the PSD obtained through LDA, including [15]: the number of detectors; the choice of measurement time; the samples preparation; and post-processing factors, such as the choice of Fraunhofer or Mie diffraction models, or the optical parameters used. An important note made by Fisher et al. [15] is that LDA requires the suspension obscuration to be within a certain range (typically 10% to 20%), and this level of obscuration requires a small amount of sample (in the order of a gram, depending of particle size). For highly heterogeneous samples, it is, therefore, important to take and use representative samples to achieve this level of obscuration. Fisher et al. [15] achieved this with soils by using a precise wet riffling methodology. When available sample amounts are limited, on the other hand, LDA is far superior to mechanical methods, as even under low obscuration conditions (e.g., 2% to 10%), reasonable results can be obtained [15]. Another important limitation of LDA is that it does not accurately represent the size distribution of irregularly shaped particles [13], such as acicular crystals or fibrous minerals. The projected cross-sectional area of a non-spherical particle averaged over all possible orientations is larger than that of a sphere with an equal volume, which, for an entire population of non-spherical particles, leads to a shift of the PSD toward a coarser size [13]. Non-spherical particles can be either a feedstock (e.g., asbestos, wollastonite) or a product (e.g., aragonite, nesquehonite) in mineral carbonation reactions, so care must be taken when interpreting the PSD for these materials. This is especially important when making comparisons to PSDs obtained from other more spheroidal materials (e.g., precipitated aragonite versus calcite), or from PSDs of pre- or post-processing samples that have different morphology and size distribution. 2.3. Complementary and Alternate Techniques (Laser Reflectance, Ultrasonic Extinction, Acoustic Spectrometry) Sieving as a particle classification method and laser diffraction analysis as a particle size distribution and average particle size determination method are a useful pairing for most applications related to mineral carbonation research. However, other methodologies are available to provide either complementary information about particle size and shape or alternate testing techniques more suitable for particular applications. An important application is in situ particle size distribution
Geosciences 2018, 8, 260
5 of 20
determination, such as in a pipe flow or stirred reactor. In situ methods allow testing of particles in their native environment, eliminating the need for sampling, filtration, drying, etc., all of which can alter particles or lead to unrepresentative sample. In situ methods also allow tracking of changes in particle size distribution (i.e., kinetics) while particles react, dissolve, precipitate, crystallize, agglomerate, de-agglomerate, fragment, etc. A key requirement of in situ PSD determination is that the particle suspension cannot be diluted for analysis, hence, making the application of laser diffraction more difficult than some alternate techniques briefly discussed hereafter. Optional requirements, especially for mineral carbonation research, may be operability under elevated pressures (e.g., 100–10,000 kPa) Geosciences 2018, 8, x FOR PEER REVIEW 5 of 20 and temperatures (e.g., 0–200 ◦ C). diffraction more difficult than some alternate techniques briefly discussed hereafter. Optional requirements, especially for mineral carbonation research, may be operability under elevated 2.3.1. Laser Reflectance pressures (e.g., 100–10,000 kPa) and temperatures (e.g., 0–200 °C).
Focused beam reflectance measurement (FBRM) is a technique developed by Lasentec, now part 2.3.1. Laser Reflectance of Mettler Toledo GmbH, which provides real-time measurement of the number and dimensions beam measurement (FBRM) by is a the technique developedprobe by Lasentec, now1), part of particles inFocused situ [18]. Itsreflectance in situ capability is enabled measurement (Figure which can of Mettler Toledo GmbH, which provides real-time measurement of the number and dimensions of speed be immersed in process fluids within conduits or vessels. Here, a laser beam rotates at a high particles in situ [18]. Its in situ capability is enabled by the measurement probe (Figure 1), which can and propagates into the particle suspension to be monitored. The presence of particles in the fluid be immersed in process fluids within conduits or vessels. Here, a laser beam rotates at a high speed is detected a focused laser beam strikestoa be particle andThe backscatters the in probe. Asisthe laser andwhen propagates into the particle suspension monitored. presence of into particles the fluid moves, adetected “cord length” is measured for the duration of and the backscatters backscattering. The length of this when a focused laser beam strikes a particle into the probe. As the laser scanned moves, a “cord length” is measured foreach the duration of it thefinds, backscattering. The length ofinto this scanned chord is recorded by the instrument for particle and is transferred a chord length chord is recorded by the instrument for each particle it finds, and is transferred into a chord length distribution (CLD). In this way, CLD provides online particle count and particle dimension information. distribution (CLD). In this way, CLD provides online particle count and particle dimension A CLD can be measured at every two seconds to track the dynamics of the process, multiple probes information. A CLD can be measured at every two seconds to track the dynamics of the process, can be placed inprobes the same thesystem, probeand can temperatures from −70 to multiple can bereactor placed insystem, the sameand reactor thewithstand probe can withstand temperatures ◦ from −70 to 180 °C [18]. 180 C [18].
Figure 1.drawing Schematic of probe an FBRM [19]. with Re-used with permission from Elsevier Figure 1. Schematic of drawing an FBRM [19].probe Re-used permission from Elsevier (4383110764877). (4383110764877).
2.3.2. Ultrasonic Extinction/Acoustic Spectrometry 2.3.2. Ultrasonic Extinction/Acoustic Spectrometry technique real-timein-line in-line (i.e., in in situ) particle size distribution determination, in AnotherAnother useful useful technique for for real-time (i.e., situ) particle size distribution determination, addition to solids loading (i.e., particle concentration) measurement, is known as ultrasonic extinction in addition to solids loading (i.e., particle concentration) measurement, is known as ultrasonic spectroscopy (UES), or acoustic spectrometry. As with FBRM, these techniques are also designed for extinction spectroscopy (UES), or acoustic spectrometry. As with FBRM, these techniques are also applications where elevated pressures and temperatures are used within pipes or vessels, in addition designedtofor applications where and temperatures are used within pipes or vessels, having high tolerance for elevated aggressivepressures chemical environments. UES is capable of detecting particles in the diameterhigh rangetolerance of 0.01 μmfor to aggressive 3000 μm, with solids concentrations up toUES 70 vol %, in opaque in addition to having chemical environments. is capable of detecting emulsions, andofmeasures UE, which, as a result,up enables particles liquids in the and diameter range 0.01 µmthetofrequency-dependent 3000 µm, with solids concentrations to 70 vol %, calculation of both particle size distribution and particle concentration [20]. The principle of operation in opaque liquids and emulsions, and measures the frequency-dependent UE, which, as a result, is as follows: particles smaller than the acoustic wavelength are detected by attenuation of ultrasonic enables calculation of particles both particle size distribution andacoustic particle concentration [20]. The principle intensity, while with diameter greater than the wavelength scatter the ultrasonic of operation is as follows: particles smaller than the acoustic wavelength are detected by attenuation waves. Extinction measurements are performed at multiple ultrasonic frequencies (i.e., wavelengths), of ultrasonic intensity, while particles with diameter greater than the acoustic wavelength scatter the and models are used to describe the viscous and thermal losses due to partial entrainment of the ultrasonic waves. particles by the ultrasonic wave, thus generating an attenuation distribution as a function of particle Extinction measurements are performed at multiple ultrasonic frequencies (i.e., wavelengths), size, which provides both the PSD, as well as the solids concentration [20]. It is also possible to and models are used to describe the viscous and thermal losses dueevaluating to partialthe entrainment differentiate spherical and arbitrary shaped particles by empirically extinction of the particles function. by the ultrasonic wave, thus generating an attenuation distribution as a function of particle Geer and Witt [20] suggest that the particle size distribution be first measured by LDA, and the results be then mathematically combined with the measured ultrasonic extinction spectrum of the same sample. The aim of this kind of calibration procedure is to evaluate the extinction function
Geosciences 2018, 8, 260
6 of 20
size, which provides both the PSD, as well as the solids concentration [20]. It is also possible to differentiate spherical and arbitrary shaped particles by empirically evaluating the extinction function. Geer and Witt [20] suggest that the particle size distribution be first measured by LDA, and the results be then mathematically combined with the measured ultrasonic extinction spectrum of the same sample. The aim of this kind of calibration procedure is to evaluate the extinction function only within the wave number limits given by the minimum and maximum particle size determined by LDA. An important limitation of acoustic techniques is the relatively long spectrum acquisition, which is in the order of several minutes, which causes difficulties in real-time monitoring of PSD, such as when the flowing solids loading level fluctuates frequently [21]. DosRamos [21] also makes an important observation regarding real-time PSD monitoring for process control purposes, which is that PSD curves are not practical as set points. In turn, calculated mean particle diameter values (by volume, weight, surface area, or number) can be used as set-points, but can also be misleading since a single mean particle diameter (e.g., D(4,3) = 100 µm) can be produced by an infinite number of PSD curves, so product properties can change more quickly and more significantly than variations in D-values might suggest. This same caution is very applicable to mineral carbonation research, where authors can misinterpret their own results or the results of others when using mean D-values as a basis for discussion and even modeling. DosRamos [21] suggests that surface area average sizing data (i.e., D(3,2)) are less prone to sharp fluctuations from minor populations, and similarly percentile figures such as the D90 and, thus, can be more suitable as a set point for process control. 3. Discussion of PSD Utilization in Mineral Carbonation Research 3.1. Use of Particle Size Distribution In Carbonation Research of Santos et al. In this section, mechanistic insights into mineral carbonation reactions, obtained through determination of PSD, are reviewed from three comprehensive studies published by Santos et al. [22–24] in recent years. The volume-based particle size distributions and the average particle diameters presented in this section have been determined by wet LDA (Malvern Mastersizer S) in sonicated deionized water, with a detection range of 0.06–878.7 µm. 3.1.1. Ultrasound-Intensified Mineral Carbonation Ultrasound has been investigated as a way to promote particle breakage during slurry carbonation, and to remove the carbonated shell or depleted matrix layers that surround the unreacted particle core, thus reducing diffusion limitations and exposing unreacted material to the aqueous phase [22]. Sonication (Hielscher UP200S, 24 kHz, 200 W, Hielscher Ultrasonics GmbH, Teltow, Germany) was shown to increase the conversion (94% vs. 86% over 25 min) and the process kinetics (0.65 vs. 0.59 g-CaCO3 /min) of calcium hydroxide carbonation, compared to mechanical stirring. Furthermore, particle size reduction of calcium hydroxide and carbonate powders is significantly greater than that of steel slag particles over 30 min of sonication (Table 1). This is likely a result of lower hardness, confirming the premise that sonication can clear the surface of unreacted mineral particles from the rate-limiting coverage of a carbonated shell. Table 1. Particle size reduction by sonication, expressed as Sauter mean diameter [11]. Re-used with permission from John Wiley and Sons (4366260200646). Powder
Original D(3,2) (µm)
Sonicated D(3,2) (µm)
Reduction (%)
CaCO3
